jeffrey.kissel@LIGO.ORG - posted 16:03, Wednesday 17 July 2024 - last comment - 11:09, Thursday 01 August 2024(79198)
H1SUSOFI -- EUL2OSEM Matrices Restored -- Damping Loops Made Function (But Drastically Overdamped) -- Suspension Dynamics Health is Great
J. Kissel
Executive Summary: In prep for the upcoming vent, I've measured the undamped dynamical transfer functions of the H1SUSOFI. The dynamics look healthy.
The rest of the story: The OFI's digital control system has had a sordid history (mostly the OSEM2EUL and EUL2OSEM matrices, but also the damping loops), and assessment of it's health has been person power limited over the years. I've fixed all that today, and restored the suspension's control system to good health -- such that the transfer functions match the model as they did (once) in 2023. More details explaining the plots later, but we're good to go for the vent.
Today's data lives here
/ligo/svncommon/SusSVN/sus/trunk/OFIS/H1/OFI/SAGM1/Data
2024-07-17_1741_H1SUSOFI_M1_WhiteNoise_L_0p02to50Hz.xml % undamped
2024-07-17_1741_H1SUSOFI_M1_WhiteNoise_T_0p02to50Hz.xml
2024-07-17_1741_H1SUSOFI_M1_WhiteNoise_Y_0p02to50Hz.xml
2024-07-17_1950_H1SUSOFI_M1_WhiteNoise_L_0p02to50Hz.xml % damped
2024-07-17_1950_H1SUSOFI_M1_WhiteNoise_T_0p02to50Hz.xml
2024-07-17_1950_H1SUSOFI_M1_WhiteNoise_Y_0p02to50Hz.xml
Non-image files attached to this report
Comments related to this report
The OSEM2EUL and EUL2OSEM matrices were generated in a Matlab workspace with
/ligo/svncommon/SusSVN/sus/trunk/OFIS/Common/MatlabTools
make_susofis_projections.m [rev 10980]
which, notably is an improved version of the script with the Yaw lever arm bug fixed (see LHO:61353).
and populated into EPICs with
/ligo/svncommon/SusSVN/sus/trunk/Common/MatlabTools/
fill_matrix_values.m [rev 11862]
The first attached screenshot shows the "new" restored matrix values, and that they were accepted into SDF.
(Recall, the OFI lives on the front-end model "h1susifoout", and this model's safe and OBSERVE .snap files both point to the same file, so there's no need to accept the SDF DIFFs in both SAFE and OBSERVE,
$ pwd
/opt/rtcds/lho/h1/target/h1susifoout/h1susifooutepics/burt
$ ls -l
lrwxrwxrwx 1 controls 69 Oct 10 2022 OBSERVE.snap -> /opt/rtcds/userapps/release/sus/h1/burtfiles/h1susifoout_observe.snap
lrwxrwxrwx 1 controls 69 Oct 10 2022 safe.snap -> /opt/rtcds/userapps/release/sus/h1/burtfiles/h1susifoout_observe.snap
)
For the record, the previous matrix values -- where matrix elements are single DOF to DOF unity maps with gains of 1.0 were accepted into SDF on Tuesday April 25 2023, two months before the start of O4 (LHO:68984). These elements were leftovers from some guest-star testing that was done just prior -- see e.g. LHO:68910, which was motivated by evidence of back scatter in the OFI, see e.g. LHO:68846.
Images attached to this comment
The data in the main aLOG above (LHO:79198) was exported and individually processed with /ligo/svncommon/SusSVN/sus/trunk/OFIS/Common/MatlabTools/ plotOFIS_dtttfs_M1.m [rev 11875] I also exported and individually processed the large back-log of data taken for the OFI since 2019, '2021-10-11_2300' | LHO:60211| Oct-Nov 2021 Vent pre-install of OFI shroud '2021-11-09_2217' | LHO:60603| Post-OFI shroud install, mechanical interference alleviated '2021-11-10_1750' | LHO:60610| Doors on, in-air close out measurements '2021-12-14_1400' | LHO:61055| Health-check post vent after pumpdown '2023-04-18_1800' | LHO:68846| Health-check in-vac, due to suspicions of backscatter caused by the OFI '2024-04-24_1500' | LHO:77391| Health-check after first major alignment excursion in O4b due to KTP crystal damage (LHO:77392); has bad basis change matrices leftover from LHO:68910 poorly restored via trending '2024-07-17_1741' | LHO:79198 | Pre-vent TFs after updating / restoring bases matrices These data are committed to /ligo/svncommon/SusSVN/sus/trunk/OFIS/H1/OFI/SAGM1/Results/. I then added the above historical log information to the TF comparison script for the OFI (after updating to absorb similar updates from LLO) /ligo/svncommon/SusSVN/sus/trunk/OFIS/Common/MatlabTools/ plotallofis_tfs_M1.m [rev 11902] such that I could make the comparison of all of this data posted as the second attachment to the main entry. The YAW DOF in those plots (page 3) exposes the history the best: 2019-10-11 17:50 -- This is the former "high loss" OFI from aLIGO, prior to the complete rebuild and upgrade that was needed for frequency-dependent-squeezing. 2021-10-11 23:00 --\ 2021-12-14 14:00 -- -- These two 2021 data sets show the OFI *after* upgrading to the FDS capable ultra low loss faraday. Apparently the yaw moment of inertia is a bit less, or the yaw cabling stiffness is a bit higher, so the resonant frequency shifts up a bit. However, this data still has the lever arm bug, so you see the magnitude is a factor of 2 larger than the model. 2023-04-18 18:00 -- This is just after I fixed the lever arm bug, and thus now the magnitude lines up with the model. It's also just prior to when the EUL2OSEM and OSEM2EUL matrices where left in the testing configuration just before and throughout O4 up until Jul 17th. 2024-04-24 15:00 -- Due to suspicions of back-scatter, with too little person power, we restored the basis change matrices to *something* but definitely not the correct basis matrices. We green-lighted the dynamics with a hand waive of "well, it doesn't look like it's *rubbing,* and we guess that it's looked like this and didn't have damping loops on for all of O4 so... probably fine?" 2024-07-17 17:41 -- This aLOG series, with the basis change matrices restored -- and the dynamics restore to the 2023-04-18 goodness.